Tech News Roundup: Christian Phone Network, AI Interpretability Tool, NSF Cuts, and China's Open-Source AI Strategy

<p>This week brings a mix of developments across technology and science, from a controversial new phone network in the US to advances in AI transparency, and from federal research funding upheaval to shifting dynamics in global AI model development. Below, we break down four key stories that are shaping the landscape.</p> <h2 id="christian-phone-network">A Controversial Christian Phone Network Blocks Porn and Gender Content</h2> <p>A new US-wide cellular network targeting Christian consumers is set to launch next week. The service employs network-level filtering to block access to pornography, and this filter cannot be disabled—even by account holders who are adults. Additionally, the company is rolling out an optional filter aimed at blocking sexual content related to gender and transgender issues. This filter is turned on by default across all plans.</p><figure style="margin:20px 0"><img src="https://wp.technologyreview.com/wp-content/uploads/2026/05/radiant-phone2.webp?resize=1200,600" alt="Tech News Roundup: Christian Phone Network, AI Interpretability Tool, NSF Cuts, and China's Open-Source AI Strategy" style="width:100%;height:auto;border-radius:8px" loading="lazy"><figcaption style="font-size:12px;color:#666;margin-top:5px">Source: www.technologyreview.com</figcaption></figure> <p>The approach raises concerns about overreach. Many websites do not fit neatly into single categories, leaving the founder—a self-described maverick—with broad, subjective control over what is allowed or banned. Critics worry this could lead to arbitrary censorship beyond the stated religious values. For deeper analysis, <a href="#full-story-1">read the full story</a>.</p> <h2 id="debugging-llms">Debugging AI Models with Mechanistic Interpretability</h2> <p>San Francisco–based startup Goodfire has released a new tool called Silico that allows researchers to peer inside an AI model and adjust its parameters during training. The goal is to make building AI models less like alchemy and more like a science. Silico uses a technique known as mechanistic interpretability to map the neurons and pathways inside a model, enabling developers to tweak them to reduce unwanted behaviors or steer outputs.</p> <p>By exposing what the company calls the “knobs and dials” of a model, Goodfire hopes to bring AI training closer to traditional software engineering. This tool could give users more control over how AI technology is built than was once thought possible. For more details, <a href="#full-story-2">read the full story</a>.</p> <h2 id="nsf-firings">Mass Firings at NSF Deal a Fresh Blow to American Science</h2> <p>The National Science Foundation (NSF) has been dealt another major setback. On the past Friday, all 22 scientists overseeing the agency’s research portfolio were terminated. The NSF is a federal agency that funds major research projects at around $9 billion annually.</p><figure style="margin:20px 0"><img src="https://wp.technologyreview.com/wp-content/uploads/2024/08/Mountain_Pass_thumb.jpg?w=3000" alt="Tech News Roundup: Christian Phone Network, AI Interpretability Tool, NSF Cuts, and China's Open-Source AI Strategy" style="width:100%;height:auto;border-radius:8px" loading="lazy"><figcaption style="font-size:12px;color:#666;margin-top:5px">Source: www.technologyreview.com</figcaption></figure> <p>Since 2025, the NSF has faced budget cuts, grant terminations, and mass firings. Staff numbers have dropped sharply, and many ambitious projects have ground to a halt. These changes represent a significant shift in how American science is funded and governed. Researchers and policymakers are now assessing what this means for the future of U.S. scientific leadership. <a href="#full-story-3">Read the full story</a> for analysis on what comes next.</p> <h2 id="china-open-source">China’s Open-Source AI Bet Challenges Silicon Valley</h2> <p>Silicon Valley AI companies typically follow a familiar playbook: keep models behind an API and charge for access. China’s leading AI labs are playing a different game, releasing “open-weight” models that developers can download, adapt, and run on their own hardware.</p> <p>This approach went mainstream after DeepSeek open-sourced its R1 model, which matched top US systems at a fraction of the cost. Beyond technical performance, the move also won something subtler: goodwill with developers. A growing cohort of Chinese companies is now embracing open-source distribution, potentially reshaping the global AI landscape. <a href="#full-story-4">Read the full story</a> for more insight into the implications.</p> <p><em>These stories first appeared in MIT Technology Review’s newsletters. To receive updates like these in your inbox, consider subscribing to The Download or The Checkup.</em></p>  <p id="full-story-1" style="display:none;">Full story placeholder</p> <p id="full-story-2" style="display:none;">Full story placeholder</p> <p id="full-story-3" style="display:none;">Full story placeholder</p> <p id="full-story-4" style="display:none;">Full story placeholder</p>